Yagis, Ekin and Workalemahu Atnafu, Selamawet and Garcia Seco De Herrera, Alba and Marzi, Chiara and Giannelli, Marco and Tessa, Carlo and Citi, Luca and Diciotti, Stefano (2021) Deep Learning in Neuroimaging: Effect of Data Leakage in Cross-validation Using 2D Convolutional Neural Networks. Scientific Reports, 11 (1). 22544-. DOI https://doi.org/10.1038/s41598-021-01681-w
Yagis, Ekin and Workalemahu Atnafu, Selamawet and Garcia Seco De Herrera, Alba and Marzi, Chiara and Giannelli, Marco and Tessa, Carlo and Citi, Luca and Diciotti, Stefano (2021) Deep Learning in Neuroimaging: Effect of Data Leakage in Cross-validation Using 2D Convolutional Neural Networks. Scientific Reports, 11 (1). 22544-. DOI https://doi.org/10.1038/s41598-021-01681-w
Yagis, Ekin and Workalemahu Atnafu, Selamawet and Garcia Seco De Herrera, Alba and Marzi, Chiara and Giannelli, Marco and Tessa, Carlo and Citi, Luca and Diciotti, Stefano (2021) Deep Learning in Neuroimaging: Effect of Data Leakage in Cross-validation Using 2D Convolutional Neural Networks. Scientific Reports, 11 (1). 22544-. DOI https://doi.org/10.1038/s41598-021-01681-w
Abstract
In recent years, 2D convolutional neural networks (CNNs) have been extensively used to diagnose neurological diseases from magnetic resonance imaging (MRI) data due to their potential to discern subtle and intricate patterns. Despite the high performances reported in numerous studies, developing CNN models with good generalization abilities is still a challenging task due to possible data leakage introduced during cross-validation (CV). In this study, we quantitatively assessed the effect of a data leakage caused by 3D MRI data splitting based on a 2D slice-level using three 2D CNN models to classify patients with Alzheimer’s disease (AD) and Parkinson’s disease (PD). Our experiments showed that slice-level CV erroneously boosted the average slice level accuracy on the test set by 30% on Open Access Series of Imaging Studies (OASIS), 29% on Alzheimer’s Disease Neuroimaging Initiative (ADNI), 48% on Parkinson’s Progression Markers Initiative (PPMI) and 55% on a local de-novo PD Versilia dataset. Further tests on a randomly labeled OASIS-derived dataset produced about 96% of (erroneous) accuracy (slice-level split) and 50% accuracy (subject-level split), as expected from a randomized experiment. Overall, the extent of the effect of an erroneous slice-based CV is severe, especially for small datasets.
Item Type: | Article |
---|---|
Uncontrolled Keywords: | Brain; Humans; Parkinson Disease; Alzheimer Disease; Image Interpretation, Computer-Assisted; Magnetic Resonance Imaging; Case-Control Studies; Cross-Sectional Studies; Reproducibility of Results; Predictive Value of Tests; Aged; Aged, 80 and over; Middle Aged; Female; Male; Neuroimaging; Deep Learning; Neural Networks, Computer |
Divisions: | Faculty of Science and Health Faculty of Social Sciences Faculty of Science and Health > Computer Science and Electronic Engineering, School of Faculty of Social Sciences > Institute for Social and Economic Research |
SWORD Depositor: | Unnamed user with email elements@essex.ac.uk |
Depositing User: | Unnamed user with email elements@essex.ac.uk |
Date Deposited: | 13 Dec 2021 15:12 |
Last Modified: | 07 Aug 2024 19:30 |
URI: | http://repository.essex.ac.uk/id/eprint/31887 |
Available files
Filename: s41598-021-01681-w.pdf
Licence: Creative Commons: Attribution 3.0